Machine Learning Workflow - ignacio-alorre/Data-Science GitHub Wiki

Workflow to follow in Applied Machine Learning

• 1 - Problem Definition

  • Problem Description
    • Informal description
    • Formal description
    • Assumptions
  • Provided Data
    • Constrains imposed on data
    • Attribute definition
  • Motivation
    • Motivation
    • Benefits
    • Use
  • Manual Solution

• 2 - Analyze Data

  • Summarize Data
    • Data Structure
    • Data Distribution
  • Visualize Data
    • Attribute Histograms
    • Pairwise scatterplots of attributes

• 3 - Prepare Data

  • Select Data
  • Preprocess Data
    • Formatting
    • Cleaning
    • Sampling
  • Transform Data
    • Scaling
    • Decompositon
    • Aggregation

• 4 - Evaluate Algorithms

  • Test Harness and Options
  • Explire and select algorithms
  • Interpret and report results

• 5 - Improve Results

  • Algorithm Tuning
  • Ensemble methods
    • Bagging
    • Boosting
    • Blending
  • Extreme Feature Engineering

• 6 - Present Results

  • Present Results
    • Context
    • Problem
    • Solution
    • Findings
    • Limitations
    • Conclusions
  • Operationalize Algorithm